aiqbench analysis
This research introduces Normalizing Flow Flow Maps (NF$^2$M), a novel architecture for Boltzmann Generators that integrates stochastic flow maps with classic normalizing flows to improve sampling efficiency in molecular systems. By reweighting denoising transitions with exact likelihoods at each step, the method avoids the computational waste and statistical unreliability common in existing generative approaches for equilibrium samples. The technique demonstrates enhanced performance on peptide systems, offering a more robust framework for generating independent configurations in computational statistical mechanics.
Bridging Stochastic Flow Maps and Boltzmann Generators with Normalizing Flows
arXiv:2609.31978v1 Announce Type: cross Abstract: Generating independent, equilibrium samples of molecular systems at scale remains a central obstacle in computational statistical mechanics. Boltzmann Generators address this by pairing a generative model with importance sampling to obtain consistent samples from the target distribution. We introduce Normalizing Flow Flow Maps (NF$^2$M), which combines the strengths of recent stochastic flow maps with the tractability of classic normalizing flows to build a Boltzmann Generator. Unlike most methods, which correct the generative model only at the end, NF$^2$M reweighs each denoising transition as generation proceeds, avoiding wasted compute on trajectories that are ultimately discarded. At each denoising step, a conditional normalizing flow proposes clean configurations given the current noisy state (a simpler task than sampling directly from the target) and its exact likelihood enables correcting each proposal toward the true denoising transition of the target Boltzmann distribution. This is in contrast to most existing methods, whose likelihoods are approximate or expensive to evaluate, undermining the statistical reliability of the correction. We establish consistency of the corrected transitions and bound how approximation errors propagate through the sampling chain. We evaluate NF$^2$M on peptide systems, demonstrating improved sampling efficiency and sample quality.
Abstract:Generating independent, equilibrium samples of molecular systems at scale remains a central obstacle in computational statistical mechanics. Boltzmann Generators address this by pairing a generative model with importance sampling to obtain consistent samples from the target distribution. We introduce Normalizing Flow Flow Maps (NF$^2$M), which combines the strengths of recent stochastic flow maps with the tractability of classic normalizing flows to build a Boltzmann Generator. Unlike most methods, which correct the generative model only at the end, NF$^2$M reweighs each denoising transition as generation proceeds, avoiding wasted compute on trajectories that are ultimately discarded. At each denoising step, a conditional normalizing flow proposes clean configurations given the current noisy state (a simpler task than sampling directly from the target) and its exact likelihood enables correcting each proposal toward the true denoising transition of the target Boltzmann distribution. This is in contrast to most existing methods, whose likelihoods are approximate or expensive to evaluate, undermining the statistical reliability of the correction. We establish consistency of the corrected transitions and bound how approximation errors propagate through the sampling chain. We evaluate NF$^2$M on peptide systems, demonstrating improved sampling efficiency and sample quality.
Machine Learning (stat.ML); Machine Learning (cs.LG)
arXiv:2609.31978 [stat.ML]
(or arXiv:2609.31978v1 [stat.ML] for this version)
https://doi.org/10.48550/arXiv.2609.31978 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 287 words
Related coverage
Self-supervised In-context Operator Learning for Stochastic Mean-Field Control
arXiv · Aug 20, 2026
SESaMo: Symmetry-Enforcing Stochastic Modulation for Normalizing Flows
arXiv · Jul 24, 2026
Presentation: Exploding GPUs
InfoQ · Oct 16, 2026
Boltzmann generators for amorphous particle systems
arXiv · Jul 24, 2026